Re: [PATCHES] Win32 CHECK_FOR_INTERRUPTS() performance

["Magnus Hagander" <mha(at)sollentuna(dot)net>]

> > Here's another version of this patch ;-) I've based it on 
> your patch, 
> > so the changes to ovalue etc should sitill be there.
> 
> In the spirit of incremental improvement ... I've taken 
> Magnus' version and added the proposed change to re-enable 
> Qingqing's patch by skipping WaitForSingleObjectEx altogether 
> in the CHECK_FOR_INTERRUPTS code path.
> I also removed WaitForSingleObjectEx in 
> pgwin32_poll_signals(), which AFAICS should be just like 
> CHECK_FOR_INTERRUPTS.  I think this is what we are proposing 
> to actually apply to 8.1beta4.  I can't test it though, so 
> please check it over...

That does seem right. I think the only reason it was there i nthe first
place was to deliver the APCs. 

But. in theory, we can get a false positive from
UNBLOCKED_SIGNAL_QUEUE(), right? Since we do it unlocked between two
threads. If we do that, we'll "recover" in dispatch_signals, because
we'l lcheck again locked and not dispatch any signals. *but*. If this
happens, we will return EINTR even when there is no signal. That doesn't
seem correct to me. It's a very small window, but it should be possible,
no?

We probably need an actual check, so for example have
dispatch_queued_signals return a value indicating if any signals were
actually dispatched, and use that to control EINTR?
 
Comments? Or am I completely off being too tired right now? ;-)

//Magnus